> -----Original Message-----
> From: linux-rdma-owner@xxxxxxxxxxxxxxx <linux-rdma-
> owner@xxxxxxxxxxxxxxx> On Behalf Of Jason Gunthorpe
> Sent: Friday, April 20, 2018 2:17 PM
> To: Steve Wise <swise@xxxxxxxxxxxxxxxxxxxxx>
> Cc: dledford@xxxxxxxxxx; leon@xxxxxxxxxx; linux-rdma@xxxxxxxxxxxxxxx
> Subject: Re: [PATCH v3 rdma-next 5/5] iw_cxgb4: dump detailed provider-
> specific QP information
>
> On Fri, Apr 20, 2018 at 02:03:22PM -0500, Steve Wise wrote:
> > >
> > > On Fri, Apr 20, 2018 at 01:32:32PM -0500, Steve Wise wrote:
> > > > > > +static int fill_sq(struct sk_buff *msg, struct t4_wq *wq)
> > > > > > +{
> > > > > > + struct nlattr *entry_attr;
> > > > > > +
> > > > > > + /* WQ+SQ */
> > > > > > + entry_attr = nla_nest_start(msg,
> > > > > RDMA_NLDEV_ATTR_PROVIDER_ENTRY);
> > > > > > + if (!entry_attr)
> > > > > > + goto err;
> > > > >
> > > > > Er, you want to nest ? But how will anything figure out what this
> > > > > RDMA_NLDEV_ATTR_PROVIDER_ENTRY is connected to? Seems like
> > > > > something
> > > > > is missing here.
> > > > >
> > > > > I think I'd rather see no nesting and have you prefix the strings
with
> > > > > something to group them??
> > > > >
> > > >
> > > > The nesting allowed for the rdma tool to take that as a time for a
new
> > line.
> > > > So you'd get the qp-global stuff for this qp on one line, then the
> > > > sq-specific on the next line, and then rq-specific.
> > >
> > > Oh gross.. no..
> > >
> > > > > > + if (rdma_nl_put_provider_u32(msg, " idx", idx))
> > > > >
> > > > > Why leading spaces? No.. Saw this a couple of times.
> > > >
> > > > Again, this made for nicer formatting on the display side. It is
kind
> > of
> > > > tricky to get decent formatting/readability with these generic
provider
> > > > attributes.
> > >
> > > Also gross, no..
> > >
> >
> > Any suggestion on how the rdma tool should display this opaque data so
it
> is
> > readable?
>
> Probably don't get to have that in an ideal sense..
>
> But have ip route list them one per line is much better than an
> unstructure jumble of many per line..
That gets terribly long. But is it better than all on one line. Hence my
grouping attributes as sets of nested that all get displayed on one line for
each nested set.
>
> Jason
What about a PRINT_FMT type attr? With an enum value like NEWLINE, INDENT,
etc?
I'm beginning to think my effort should have been spent adding what I need
to debugfs for iw_cxgb4. I'm not sure what I want applies across all rdma
providers...
Steve.
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
